Radio Propagation Models Comparison and Tuning

for LTE Based on Experimental Data in Tripoli-

Libya
Abdalraof R. Hams Nagia M. Gaboua Mostafa B. Abuitbel Ismail M. Shrena
Dept. of Electrical Engineering Dept. of Electrical and Dept. of Electrical and Dept. of Mobile Computing
and Computer Electronic Engineering Electronic Engineering Faculty of Information
Libyan Academy for Faculty of Engineering Faculty of Engineering Technology
Postgraduate Studies University of Tripoli Azzaytuna University University of Tripoli
Janzour - Libya Tripoli - Libya Tarhuna - Libya Tripoli - Libya
[email protected] [email protected] [email protected] [email protected]

Abstract— The mobile operators in Libya, like many other presented work was done on the city of Tripoli to select the
operators, use some propagation models to predict the strength best possible model and aimed to improve.
of the signal at the receiver, but without analyzing which model
is the best for the Tripoli region. This paper presents a B. Comman propagation models
comparison of four radio propagation empirical-type models Among the propagation models, empirical-type models
(COST 231, Okumura, Okumura-Hata, and Ericson 9999) being based on measuring efforts are adopted in the current
based on LTE technology to be applied at different locations work. Here are some of the well-known outdoor implemented
within Tripoli environment. The field measurements analysis models.
shows that the Okumura has the lowest RMSE value, which is
being tuned using real data collected from the local • COST 231 model
environment.
The valid frequency range of COST231 is extended to
Keywords-propagation models; LTE technology; cell [1500MHz, 2000MHz] in compare with Hata, which was
planning. [150MHz to 1500MHz] [2].

I. INTRODUCTION The COST 231 path loss is given by

In many residential areas, there is difficulty in = 46.3 + 3.9 log − 13.82 log ℎ −
communicating via mobile phone networks because it is far ℎ 44.9 − 6.55 log ℎ log + (1)
from the cells set up by companies that provide mobile phone ℎ = 3.2 11.75ℎ !
− 4.97 #$ %& $ (2)
communication services, or because of signal reception
problems or weakness in general, or because the place where = 3 #$ # '#()#* & 0 #$ * '#(#)*
the user is located has natural obstacles or tall buildings.
Therefore, it is the responsibility of the cell planning engineer Where:
to design the network based on the prediction of the signal : Frequency [MHz]
strength and to what extent it can be reached by placing the
cells accordingly. ℎ : Height of base station [meters]
Moreover, working without planning becomes a kind of ℎ : Height of mobile station [meters]
ℎ
futility and a waste of time as reaching goals becomes elusive
and its importance is evident in his expectations for the future : Correction factor of mobile antenna height.
as the goals to be reached are future goals, meaning that d : link distance [Km]
achieving them is done within a specific period of time that
may be lengthened and shortened, which imposes on a cell • Okumura model
planning engineer makes the necessary assumptions about The Okumura model was created based on measured data
what this future might be and form an idea of what the in Tokyo, Japan. It works perfect for practice in modern cities
situation will be when starting to implement the goals and but not a lot of towering structures.
during the various implementation stages.
The expression of the Okumura Model is formulated
II. PROPAGATION MODELS below:
A. Theory - . = / + 0-1 , − 3 ℎ4 − 3 ℎ5 −
There are three kinds of propagation models which are 36476 (3)
empirical (time dispersive and non-time dispersive),
Where:
stochastic and deterministic where each of these models has
advantages and disadvantages. These models are also differ in - . : Median of path loss
approach, complexity and accuracy [1]. Thus, the choice of
the radio propagation model is the first and most important / : Loss in free space
step in the work of the cell planning engineer, as the more 0-1 , : the median attenuation relative to free space
accurate the model used is, the better the network will be in
terms of coverage, by choosing the appropriate place for 3 ℎ5 : the gain factor for base station antenna height.
placing the cells, and many types of research have been made 3 ℎ4 : the gain factor for mobile antenna height
in this science area, but because each city has its nature. The
 ;:
G h: = 20 log (4) • EPRE
!
;< EPRE indicates the power for one resource element (RE).
G h< = 20 log h< <= 3m (5)
= This can be used for any channel (Reference Signal, PDSCH)
>?@A? is the type of environment gain
[2].
EPRE = total Tx power - 10Log(number of resource block
• Okumura-Hata Model (RB) x 12)
Okumura model has been modified where the effects of
city structures (scattering, diffraction, and reflection) has been • Antenna Gain:
considered. The frequency range is [150MHz, 1500MHz] and Antenna gain is function of type and size of antenna and
mobile Station Antenna Height in range of [1m,10m]. The the frequency used as carrier. Gain in cell tower antenna is
height of base station Antenna is [20m, 200m] and link typically 15-18dBi [2].
distance is up to 20km. The path loss of Okumura Hata model
is calculated using Eq.6: • Losses
= 69.55 + 26.16 − 13.82 ℎ − B
Losses involve losses on both sides of cables and body
+ C44.9 − 6.55 ℎ D 6 (cell tower antenna and phone). The cable length and type and
frequency have a major effect on the cable losses. The loss in
For medium-sized or even small or city: the cable of the cell tower antenna varies from 1dB to 6dB. In
B = 0.8 + 1.1 − 0.7 ℎE planning, the phone lose is considered to be 0dB.
− 1.56 7 • Effective Isotropic Radiated Power (EIRP)
Where: EIRP define the value of transmitted power radiated in the
: G (ℎ ** ) # $ C .D desired direction including the gain of the transmitter antenna
and the losses [2]:
: Frequency [MHz]
EIRP (dBm) = Pt (dBm) + Ga (dBi) – Cable losses (9)
ℎ : Height of base station [meters]
Where:
ℎ : Height of mobile station [meters]
Pt(dBm): transmitting power of cell tower
B : Correction factor of antenna height
Ga(dBi): Gain of Antenna in comparison to isotropic
d : link distance in [Km] antenna.
• Ericson 9999 Model • RX Level
The Ericsson 9999 Model is a software tool for wireless The RX level is a received signal level considering the
network planners provided by the Ericsson Company. This signal strength obtained from the cell tower antenna by the
model was developed to use in different propagation mobile phone. It is importing factor which determines whether
environments [2]. Ericsson propagation model equation the reception was good or not [3]. The mathematical formula
illustrated in Eq.8, where the parameters k0, k1, k2, and k3 to calculate the RX level is:
depends on the environment type as shown in table I.
RxLev(dBm)=EIRP-Pl – L + G (10)
= H + H + H! ℎ + H= ℎ .
− 3.2C 11.75 ! D + 44.49 Where:.
− 478C D! 8 Pl(dB) = path loss in propagation model.
C. Radio Link Budget Calculation L(dB) = margin of (low fading, Interference) and loss
The calculation of link budget is based on farthest point (body, penetration)
from the cell tower [2]. The minimum signal strength received G(dB) = other gain (pilot boosting gain, MIMO gain,
by the receiver must be considered in determine the maximum Handoff. Gain, HARQ gain and coding gain) [3].
coverage. The important parameters of the budget calculation
are discussed in the following definition: The reference signal receiver power (RSRP) has been
measured to represent experimental receiver power, which is
• Transmitter Power an averaged value for all the resource elements for reference
signal within a symbol. Since this is the averaged value, the
Typical value of the total transmission power from cell
value would be similar to the energy per resource element
towers antenna is from 36dBm to 46dBm [2]. (EPRE) value that was set for the reference signal [4].
TABLE I: ERICSON MODELS PARAMETERS III. MODEL ASSESSMENT METHOD
Environment Type IJ IK IL IM Well regression model offers the possibility to minimize
Rural 45.95 100.6 12 0.1 the difference between predicted and observed data values.
Suburban 43.20 68.63 12 0.1
Urban 36.20 30.20 12 0.1
The Root Mean Square Error ((RMSE)) was used in this work
to measure the accuracy of the propagation models.
A. The RMSE
The RMSE is a method frequently used for error analysis.
It precisely shows the magnitude of deviations and estimated
(measured or calculated) value from an actual value. RMSE is
analytically given by Eq.11 [5]:
X ∑ UVWU< X
RMSE = RS Z (11)
YW

Where:
Pm: PL (Measured [dB])
Pr: PL from the models (calculated [dB])
N: Number of data points (measured).
IV. MODEL TUNING METHOD
Regression analysis is one of the most common statistical
techniques practically used in various applications such as
engineering and others [6]. Simple linear regression has been
used for the current model tuning which is a model with a
single independent variable x1,x2,…xn that has a straight-line
relationship with another dependent variable y. Apply the
values of x in the relation to estimating the value of variable
y. To make the causal relationship more effective and accurate
in the calculations, cover variables x1, x2, … xn to
characterize the response variable y. This will allow finding
which independent variables x1,x2,…xn are more useful than
others. The detailed procedures for performing the regression
analysis are shown in Fig. 1.
It is useful to display x as determined by the data analyzer
and measured with little error, while the y response is a
random variable. That is, there is more than one value of y at
each possible value of x. This distribution means:
y=β0+β1x (12) Fig. 1: Procedures for performing the regression analysis

V. RESULTS AND DISCUSSION TABLE II: LOCATIONS OF THE CELL TOWERS

Experimental measurements of radio propagation models
Cell name Long. Lat.
are done for three different locations. Practically, the strength
of the signal is measured by a three cells drive test in Tripoli NT16 13.132925 32.871725
city from Almadar Aljadid network Libya. These cells have NT500 13.122151 32.849007
been chosen in a building environment with a maximum tall
of 10m available in Tripoli. Table II shows the locations of the NT179 13.130207 32.858314
cell towers (urban), which is used for driving test. The
parameters, which are used to run the investigated models, are TABLE III: PROPAGATION MODEL PARAMETERS
illustrated in table III.
Parameter Value
Furthermore, the following software are used to achieve Frequency (MHz) 1855
the proposed work and present the obtained results: Total Tx Power (dBm) 43
EPRE using eq.(7) (dBm) 12.2
i. Matlab was used in all calculations and plots of EIRP using eq. 28.7
the utilized formulas. Allocated RB 100
ii. TEMS discovery releases 20.3.0 has been used to Tx Antenna Gain (dBi) 17
Tx Cable Loss (dB) 0.5
perform drive tests [7]. Tx Body loss (dB) 0
A. Receiver signal strength calculation Rx Antenna Gain (dBi) 0
Interference Margin (dB) 3.67
Using Eq. 10 with input data from table 3, the strength of Penetration Loss (dB) 18
the signal for each model can be computed using the following Tx Antenna Height (m) 22
equation: UE Antenna Height (m) 1.5

[\ . = 11.03 − G
Slow Fading Margin (dB) 7
(13) Coding Gain 0
Multi-Antenna Combining Gain 3
Table 4 presents rreceiver ssignal strength (RSS) Pilot Power Boosting 3
equations using propagation loss equations illustrated before HARQ gain 3
and with input data from table III.
 Now using the propagation model to draw signal strength TABLE VI: MODIFIED OKUMURA MODEL PREDICATION VALUES
prediction in comparison with real measurement in range up ^ Value
to 600m as shown in Fig. 2. ^J -27.978580693942
^K 0.615528733540788
B. Models assessment using RMSE

RMSE has been calculated for each model and listed in

table V. From the table it appears that Okumura has the lowest
RMSE value than the other models.
C. Okumura model tuning
This section illustrates a description of how to use the
linear regression method to modify the Okumura model to
match the Tripoli environment.

y=β0+β1x (14)

The value of y represents the predicted RSS and to have

the accurate value of the prediction, ] and ] are given in
table VI.

After applying the new equation, a value of RMSE was

calculated and it was 3.8811. Referring to table 5, the lowest Fig.3. Receiver signal strength after tuning vs measured data
RMSE value for the Okumura model was 6.2754 so the
accuracy of prediction in cell planning increased by almost VI. CONCLUSION
double. Fig. 3 illustrates the Okumura after tuning in This paper was focused on studying the propagation
comparison with Okumura before tuning with real measured models to be used in cell planning in the Tripoli-Libya
data in rang up to 600m. environment. The comparative analysis indicates that the
TABLE IV: RSS EQUATIONS Okumura model has the lowest RMSE value than the other
Model Path loss eq. RxLev. eq.
models utilizing real measurements. The tuning of the
Okumura model using linear regression methods to improve
Okumura 110.3631 + 20log −99.331 − 20log
the Okumura model was achieved by using the root mean
Okumura-Hata 136.15 + 36.107log −125.1225 − 36.1 log square error (RMSE) test to provide sufficient RMSE for
Ericsson 9999 158.5522 −147.522 − 69.06log radio prediction and utilize the network effectively. Thus, the
+ 69.06424log tuned Okumura was a promising model to fulfil the desired
Cost 231 138.544578 −127.514 − 36.1log objective.
+ 36.107log
VII. REFERENCES
1. M. S. Mollel and M. Kisangiri, “An Overview of Various Propagation
Model for Mobile Communication”, PACT, July 2014, pp. 148-153.
2. K. H. Haider; A. Intisar and I. J. Abbas, "Analyzing Study of Path loss
Propagation Models in Wireless Communications at 0.8 GHz", Journal
of Physics: Conference Series. Vol.1003. No.1. IOP Publishing, 2018,
pp. 1-8.
3. J. Liu, C. Tangbai, and Y. Bo", Long Term Evolution (LTE) Radio
Access Network Planning Guide", China: Huawei Technologies CO,
2011.
4. https://www.sharetechnote.com/html/Handbook_LTE_RSRP_EPRE_
TotalPower.html 18-12-2019
5. Neill, Simon P. and M. R. Hashemi, “Fundamentals of Ocean
Renewable Energy: Generating Electricity from the Sea”, Academic
Press, 2018.
6. X. Yan and X. Su.,”Linear Regression Analysis: Theory and
Computing”, World Scientific, 2009.
7. https://www.infovista.com/tems/discovery 16-12-2019.

Fig.2. Receiver signal strength prediction vs measured data

TABLE V: RMSE FOR PROPAGATION MODELS

Model RMSE RMSE RMSE
NT16 NT179 NT500
Okumura 8.21507922 6.275391287 9.5416575
Okumura-
22.3468849 19.72076182 21.7371217
Hata
Ericsson
29.06547967 22.86805954 27.6544103
9999
Cost 231 24.426969721 21.955186248 23.6809099